As we get closer to the election I’ll be writing about what the polling/data is telling us more regularly. But as we’re still probably ten months away I am going to stick to monthly round-ups for now.
In this one I’m going to:
Dig into the details of *that* YouGov MRP
Cover the key points on the new constituency boundaries
Discuss the biggest methodological question for pollsters: what to do with “don’t knows”
List other interesting datapoints and analyses of note.
Clearly the overall picture is that we’re on track for a Labour majority and possibly a landslide. What’s interesting is this has only recently been accepted as the base expectation by most of the media, despite nothing really changing over the past year. It’s almost as if moving into the calendar year in which the election will be held has jolted everyone into recognition that it’s real. Narrative shifts really matter in Westminster because they change the mood (as we’ve seen with Simon Clarke’s rather desperate kamikaze attack). The Tories are in danger of becoming so deflated, and so lacking in any reason for hope, that large numbers of them give up. It’s not impossible we may start to see defections, or more MPs like Chris Skidmore escaping to a new job before the deluge.
The YouGov MRP
This MRP – which showed a Labour majority of 120 – seems to have played a role in waking up Tory MPs and friendly media to the scale of their woes. If you’re wondering why there was so much fuss about this particular polling exercise it’s partly because YouGov have a very good track record. In 2017, when they launched their MRP for the first time, it was widely mocked for predicting Corbyn-led Labour would win seats like Canterbury. But they did and YouGov’s reputation shot up as a result. It was also pretty accurate in 2019 and a similar exercise for last year’s elections in Spain was almost spot on too.
It's harder to say why they are so good. With normal polling it’s possible to identify the key methodological differences between pollsters but not with MRPs as they are more complex and opaque. Essentially they involve taking a large sample (14k in this case) and breaking it down by demographic group, then assessing the demographic breakdown in each constituency to make a prediction. But we don’t know the exact mix of demographics they use, the importance they place on them, or, critically, how they take constituency-level effects like tactical voting into consideration. We do know that the bigger the sample the more likely they are to have enough information to make useful inferences about different groups and seats. A MRP with fewer than 10k respondents is not worth bothering with.
But even with similar sized samples, methodology makes a significant difference. For instance in November a MPR by the pollster Survation, which had similar topline voting intention numbers, had a seat count of:
Labour 434
Conservative 151
LD 15
SNP 28
YouGov though had:
Labour 385
Conservative 169
LD 48
SNP 25
So what explains the difference? Well the lower Labour total for YouGov can be partly explained by their decision to allocate undecided voters to the party they best fitted, which brings the Labour vote down. The higher Lib Dem total must have something to do with the way they account for constituency-level differences: in effect they are showing a much greater level of tactical voting in Lib Dem target seats.
Taken together this means the Survation MRP was more bullish for Labour – showing them winning extremely longshot seats like John Major’s old Huntingdon constituency, which YouGov does not. But conversely YouGov is more bullish for the Lib Dems, showing them, for instance, winning the new Henley and Thame seat in Oxfordshire, whereas Survation had them coming third.
For me the YouGov MRP passes the sniff test better. The results just look more plausible. And they do have the track record. But given we can’t see the detailed methodology it’s all a bit of a leap of faith.
If we assume the YouGov version is a reasonably good assessment of where things stand at the moment then digging deeper into their numbers shows just how perilous a situation the Tories are in:
Keep reading with a 7-day free trial
Subscribe to Comment is Freed to keep reading this post and get 7 days of free access to the full post archives.